Masonry walls are subjected to cyclic lateral loads to study damage caused by earthquakes. Imaging techniques are useful to quantify crack networks in such tests, where preferential locations for their initiation are not present. However, detecting small cracks in large structures is challenging. It is shown that FE-based digital image correlation (DIC) can detect and quantify cracks by combining optical and mechanical information of a cyclic shear experiment performed on a full-size masonry wall. Pixel-wise gray level residuals and elementary crack opening displacement fields are the key quantities of the proposed framework. Detection criteria based on standard uncertainties guided the application of new DIC strategies (i.e., mechanical regularization, mesh adaption, and damage). Two damage regimes were quantified. Zigzagged cracks were first formed for which their opening displacements were on average less than 0.5 mm with very limited damage. They were followed by sliding shear cracks whose mean opening displacements varied between 1 and 3 mm, and damage developed in a more gradual and extended way. Such rich full-field dataset may be used for validating damage models up to full scale simulations.